We are trying to develop a multi-model custom transformer with the downstream task for Video Classification. Following the step given in Transformer Documentation for “Sharing custom models” to build our Transformer Model and later publish it in Hub. Little Curious about how can we convert our Feature Extractor that extracts features from the video, is available in Auto Feature Extractor. The intention is to have same way how we are registering our models to Auto Model in the article “Sharing custom models”
Kindly share any blog or Article that helps in clearing our understanding.